Operational Efficiency Improvement: Baseline 5 Metrics First

Key takeaways

• Capture a two-week baseline of five metrics (cycle time, rework rate, utilization, cost per delivery, first-pass yield) before changing any process.
• Attribute improvement by shipping one change at a time and comparing equal-length before/after windows with the same measurement method.
• A one-tab spreadsheet with five columns is enough capture tooling for a 5-50 person firm; cap logging at 15 minutes per session.
• Structure a 90-day roadmap so each 30-day block owns one metric, one change, and one re-measure, plus an explicit don't-do list.
• For multi-site operators, baseline each site separately and use one site as the control when testing a change.

Operational efficiency improvement only counts if you can prove it, and proof starts with a baseline captured before you touch any process. Measure five metrics for two weeks: cycle time, rework rate, utilization, cost per delivery, and first-pass yield. Then any change you make can be attributed with data instead of guessed at, and your 90-day roadmap is built on numbers you own.

Why measure a baseline before changing any process?

If you change three things at once and revenue moves, you cannot say which change did it. A baseline of five metrics captured before any process edit gives you attribution: each improvement traces to one change, not to a season or a lucky referral.

The cherry-picking problem

Without a baseline, owners pick tasks that feel productive and abandon them when nothing visibly compounds. As one owner put it, people "cherry-pick random tasks and nothing compounds." Admin drag is real: small businesses in Canada spend substantial owner time on compliance and paperwork, time Statistics Canada tracks in its surveys of business conditions and productivity. You cannot recover that time deliberately if you never counted it.

What 'proven' beats 'assumed' looks like

Proven means: hours per delivery, response time, rework rate, and margin, all recorded before the change. Assumed means "the new SOP feels faster." The first survives a board or a bank conversation; the second does not. A baseline also feeds a 90-day roadmap with targets you can actually check.

Which 5 metrics should an SME baseline first?

Capture five numbers before touching any process. Each one answers a different question, and together they cover speed, quality, capacity, cost, and reliability. If you only track one, track cycle time; if you can track two, add rework rate.

Cycle time

How long from "work accepted" to "work delivered," in business days. Measure it per service line, not averaged across everything, because a blended number hides which offer is slow. Pull it from invoices and delivery dates you already have.

Rework rate

The share of deliveries that needed a fix after the client saw them, or after your team touched them twice. Count revisions, refund requests, and "can you redo this" emails over a representative month. This is where undocumented processes usually show up first.

Utilization

Billable or delivery hours divided by available hours, per person, for a typical two-week period. In a 5-50 person firm this exposes founder-dependent work: if one person sits at 90% while three sit at 40%, the constraint is staffing shape, not effort.

Cost per delivery

Total delivery cost (labour hours times loaded rate, plus direct costs) divided by number of deliveries. Compare it to price per delivery to see which service lines actually make money. Most owner-operators have never computed this per service.

First-pass yield

The percentage of deliveries completed with no rework, no escalation, and no missed scope. It pairs with rework rate but adds the "clean handoff" test that matters for multi-site or board-facing work.

Metric	Question it answers	Where to get it	Baseline effort
Cycle time	How fast?	Invoices, delivery dates	1-2 hours
Rework rate	How often do we redo?	Email, tickets, revisions	2-3 hours
Utilization	Who is the constraint?	Timesheets or calendar	1-2 hours
Cost per delivery	What does each unit cost?	Payroll + invoices	2-4 hours
First-pass yield	How reliable?	Delivery log	1-2 hours

Run all five through the same criteria you would apply to any tool: if a metric cannot be re-measured in 30 days with the same method, it is not a baseline, it is a one-off. The company profile approach treats these five as the minimum dataset before any roadmap work, and the same honesty applies here: two weeks of spreadsheet work is the real cost, and skipping it means any later "improvement" is a guess.

How do you capture a baseline in two weeks without new software?

You need a spreadsheet, a calendar block, and a sampling rule. Most Canadian SMEs in your size band have fewer than 50 employees, so a full process-mapping exercise is overkill; a two-week sample is enough to make later claims defensible. Roughly 98% of Canadian businesses have fewer than 100 employees, which is why lightweight capture beats enterprise tooling at this stage (Innovation, Science and Economic Development Canada).

The 2-week capture sheet

One tab, five columns: date, metric, raw number, who recorded it, and any known disruption (staff absence, client deadline, holiday). Log each of your five metrics two or three times per week, not daily, because daily logging is what gets abandoned by day four. The disruption column matters most: when you attribute improvement later, you will need to show the baseline wasn't measured during an unusual week. If a metric needs a definition (what counts as "on time"?), write it in a notes cell now so the definition doesn't drift between baseline and post-change measurement.

Sampling rules for small teams

Sample the same weekdays each week, or your numbers compare a Monday-heavy week against a Friday-heavy one. Cap collection at 15 minutes per session; if capture takes longer, the metric is defined wrong. And resist improving anything during the two weeks, since a mid-baseline change poisons the comparison. If you want the baseline tied to a prioritized plan rather than sitting in a spreadsheet, a structured sprint like the OGPP Strategy Sprint turns captured numbers into a 90-day roadmap, though you can run the same exercise with the sheet alone.

How do you attribute improvement to a specific change?

You attribute improvement by isolating one change at a time and comparing a fixed before/after window, so seasonality and client mix can't quietly take the credit.

Before/after windows

Pick equal-length windows, ideally 4 weeks before and 4 weeks after the change, with nothing else shipped in between. If you changed intake in March, compare March against a comparable month, not against your busiest quarter. Log the change date, what shipped, and who was involved, so six weeks later you can still say what moved and when.

Ruling out noise

Three checks separate real gains from luck. First, hold one metric constant: if revenue jumped but headcount or client count also jumped, the efficiency claim is unproven. Second, check for one-off events: a single large client signing in the after window can distort hours-per-engagement, so note it or exclude that client. Third, re-run the same baseline capture method both times; if you measured hours differently after the change, the comparison is worthless.

If two changes shipped together and you can't separate them, say so in your notes rather than crediting both. An honest log beats a clean story. The OGPP Strategy Advisor structures this attribution step into the roadmap review, and an advisor pass helps when the numbers look ambiguous.

How does a baseline feed a 90-day roadmap?

Your five baseline numbers become the targets, not the starting trivia. Each 30-day block of the 90-day roadmap owns one metric, one change, and one re-measure at the end of the block. That structure is what separates a plan that compounds from the cherry-picking problem owners describe: random tasks get picked off and nothing accumulates. Strategy execution tools like Cascade are built around this plan-to-execution link, which is the same link your baseline has to provide.

One metric per 30-day block

Days 1-30: fix the weakest metric, usually response time or rework rate, and re-measure. Days 31-60: the second metric, informed by what the first change broke or improved. Days 61-90: the third. If a metric has not moved by its block's end, the attribution section tells you whether the change failed or the baseline was wrong.

Don't-do items

The baseline also names what you will not touch for 90 days. Write the don't-do list into the plan itself; our implementation support treats those constraints as part of the roadmap, not a footnote. Without them, a second priority always crowds in.

Where do structured AI tools fit versus raw ChatGPT or consultants?

If you have already captured your five baselines, the tooling question is narrower than it looks: you need something that turns your numbers into a constrained plan, then pressure-tests it. Raw ChatGPT can draft a 90-day roadmap from a prompt, but it will not hold you to one ICP, one channel and one core offer unless you supply the constraints yourself every session. Boutique consultants bring judgement on approvals, tone and board context, but at a price point that rarely fits a 5-50 person service business running on CAD $19/month memberships and $149 reviews.

Comparing options honestly

Structured tools sit between the two. Cascade and ClearPoint Strategy are strategy platforms with execution tracking; Upmetrics targets business planning; LucidFlow is a direct competitor in AI advisory. The differentiator is whether the tool forces tradeoffs and marks "don't do this" items, or just generates task lists people cherry-pick until nothing compounds.

Where OGPP Strategy Advisor falls short

OGPP Strategy Advisor runs the same test and fails some of it. It does not replace an expert review when a decision needs professional judgement; advisors cost extra, and sprint work starts from $2,500. It also assumes you have documented processes and at least rough KPIs to feed it. If you have neither, raw ChatGPT plus your own baseline spreadsheet is the cheaper starting draft. Where it fits is the middle case: you have the numbers, you need the constraints enforced, and you want a human pass before anything reaches your team or board.

What does this look like for care, property and multi-site operators?

Run the same five-metric baseline per site, not across the whole operation. A blended average hides which location is dragging response times or rework, so capture each site separately, then compare.

For care settings, swap generic KPIs for the ones that map to your duty of care: time from referral to first contact, visit completion rate, and documentation lag. For property or multi-site operators, use work-order turnaround, first-visit fix rate, and cost per site visit.

The attribution rule stays identical: change one process at one site, hold the others as your control, and measure the same window length. If site B improves while site A holds flat, you have evidence. If everything moves at once, you have a coincidence.

FAQ

How long should I wait after a process change before measuring results?

Wait until you have an equal-length window: four weeks after the change to match four weeks before, with nothing else shipped in between. Measuring sooner means seasonality or client mix can take credit for the movement. Log the change date and what shipped so the comparison stays defensible weeks later.

What if I don't have enough data to set a baseline for a new service?

Use the first two weeks of live deliveries as your baseline, even if that means only a handful of data points. Record the same five metrics from the first invoice, and note any disruption in the capture sheet. A small, honestly labelled baseline beats no baseline, because you still get attribution on the next change.

Can I do this with a spreadsheet or do I need KPI software?

A spreadsheet is enough. One tab with five columns (date, metric, raw number, who recorded it, disruption) captures everything a 5-50 person firm needs for a two-week baseline. The test that matters is whether you can re-measure with the same method in 30 days, which a spreadsheet handles as well as paid tooling.

How many metrics is too many for a 5-person team to track?

More than five is too many. Five metrics covering speed, quality, capacity, cost, and reliability already demand two or three logging sessions per week at 15 minutes each. If tracking pushes past that, capture gets abandoned by day four. Track the five, and only add a metric when a 30-day block needs a different lens.

Do non-profits need different efficiency metrics than for-profit SMEs?

The five-metric structure stays the same, but swap the definitions to match your mission. Cost per delivery becomes cost per client served, first-pass yield becomes completions without escalation, and utilization tracks staff plus volunteer hours. The attribution rule is identical: one change, equal before/after windows, and a re-measure using the original method.